Creating Comparability among Reliability Coefficients: The Case of Cronbach Alpha and Cohen Kappa

Abstract

Cronbach alpha and Cohen kappa were compared and found to differ along two major facets. A fourfold classification system based on these facets clarifies the double contrast and produces a common metric allowing direct comparability. A new estimator, coefficient beta, is introduced in the process and is presented as a complement to coefficient alpha in estimating the psychometric properties of test scores and ratings.